digitally controlled stepped attenuators..??

[Rochey]

[quote author="mcs"][quote author="Rochey"]PCM4204 (118db, 4 channel, 600mW ADC).[/quote]
Are you talking about the two channel PCM1804, or is it a new secret product the TI website hasn't been told about?
[/quote]

Oooooh, you're quick! :grin: :grin: :grin:
You can find details of the PCM4204 in the audio solutions guide (in my signature).
In terms of operation, it's similar to the PCM4202 (that's the 2 channel one). The PCM4204 also allows for a TDM mode where you chain 2 of them one after the other for 8 channels of 24bit 96Khz audio. (I'm not sure if the TDM mode supports 192KHz).

The PCM4204 is currently in engineering samples, rampup production will hopefully be before christmas.

Don't even ask for more info

 

[clintrubber]

from Rochey:

Peter

- cheers mate... Don't all engineers in the Netherlands work for phillips at one time or another??? *kidding*

Uhh, there's indeed quite a lot of us working for Philips, that's right.
But we definitely need most of the engineers to keep the
water out and the windmills going. :grin: :wink:

Proost,

Peter

 

[Rochey]

lololol,

does that mean you some like speaker technology to push the water out???

And maybe old rotating speaker cabinets to keep the windmills turning?

Proost!

R

 

[clintrubber]

does that mean you some like speaker technology to push the water out???

Nail on the head, speakers were used. But we moved them too fast and got complaints from the UK.

Enter MFB, Philips' Motional FeedBack system to keep those woofers in better control. It's only after that that the application for audio emerged.

Water is everything here. Air too. Our crownprince is an authority on the subject of water and even consists of 100% air himself ! :wink:

And maybe old rotating speaker cabinets to keep the windmills turning?

You've been visiting the Netherlands ! :grin:

Bye & have a good weekend,

Peter

 

[Mbira]

OK, I'm bringing this back to the surface. Mikkel-I was trying to open the pdf's on your website with no luck. Also, does anyone have a parts list by any chance-maybe with digi-key part numbers :roll: ? I finally have some time to work on this circuit!

Thanks-Joel

 

[mcs]

[quote author="Mbira"]Mikkel-I was trying to open the pdf's on your website with no luck.[/quote]
You have probebly discovered a bug in Acrobat Reader that Adobe doesn't fix for some reason (they know about it). When you open some PDF files using the Acrobat plugin the file is loaded but nothing is displayed - is that what happens?

If it is, just save the files on your harddrive and open them in Acrobat reader. They work fine here with Acrobat Reader 3.0 at least...

Best regards,

Mikkel C. Simonsen

 

[Mbira]

Got them-Thank you sir! What are the brand or types of relays that you are using? There's lots to choose from in digi-key, and I'm still pretty new and trying to figure it out.

A parts list from someone would be beautiful...

Joel

:guinness:

 

[pilo]

hey it's funny this thread bring back to surface today.

Mcs, do you think there's a way to make a log curve with 256 steps (with of course only 16 resistors and 8 relays)?
I think there's one... I'll post more info on that tomorow (but I think someone need to check my idea).

 

[PRR]

> make a log curve with 256 steps

As a practical man, I object that there is almost never a need for 256 steps of gain. You can't hear 1dB steps in music, and 2dB step switches are widely accepted. You almost never need, want, or can use more than about 60dB (plus "off") of range in a single volume control unit.

It is certainly possible to have 63 steps of 1dB with six relays, a tolerable number of resistors, and several active buffers. Ideally 7 buffers, but I think you could get away with 3. One of them can also be a gain stage, which is often associated with a volume/gain control.

You could do without buffers if you used constant-impedance attenuators. But all of these need at least 2 poles per channel. If you expect stereo, that means 4-pole relays or doubling-up on relays. And around modern gear which expects high input impedance and low output impedance, you will probably end up buffering the in or out or both. Since buffers are now cheaper than relay contacts, this plan may not be best.

> log curve with 256 steps (with of course only 16 resistors and 8 relays)?

My 2-second guess is "no". But post your plan. Either you are very clever, or someone will spot the reason why it won't work as hoped, which may lead to another plan that might work.

 

[NewYorkDave]

As a sidenote, I'd like to point out that the constant-impedance mixer controls of the "golden age" typically had 21 steps of 2dB each, with the last few steps tapered to "off." The really top-notch versions had 30 steps of 1.5 dB, again with the last few steps tapered.

It was established some years ago that a change of 2dB is "just perceptible" to the human ear.

 

[cuelist]

A couple of years ago, the late Barry Porter (of Trident fame) was kind enough to give me a copy of the schematics of remote controlled mic pre of very high quality. It was designed for Cadac and aimed for ultra high quality classical recording use.

It used a special type of Photo-MOS opto-relays (AQV series from NAIS I think) that switched in resistors. Each opto relay was driven by a transistor with a clever R-C time constant arrangement that ensure absolutely click free gain change.

And yes, it was reverse log law.

I lost the PDF with schematic but I do have a passable printout in a drawer somewhere. I could make an effort to find it and scan it if there is some interest in it. The mic amp itself is worth a look too.

It is too bad that Barry passed away - he was one of the most interesting people I have ever gotten to know. Eccentric to a point but very clever and always willing to share stories and knowledge. He was the unsung of many world class audio designs. Not as famous as Mr Neve but I dare say he was at least as clever and innovative. Loved his Chimay beer too!

 

[pilo]

http://membres.lycos.fr/blendinpulse/relay.html

here's the link to the schematic.

The idea is to get one resistor with a lot of value near R (total resistor of the pot), and another one with a lot of value near 0.

I use an excel sheet to make some test, but I'm too dumb and don't know how to use it to calculate all the 256 value to display it... so I work with a few value.
I choose for the parrallel resistor R, R/2, R/4... R/128 and 2R, 4R... 128R for the one in series.
R is the total resistor of the pot. Maybe with other resistor value it's possible to get a better curve...

Oh I forgot to say that R/128 is the MSB and R is the LSB for the resistor in parallel, and 128R is the MSB and R is the lsb for the other one...
I hope this is understable...

so where I am wrong here?

 

[Mbira]

So if I wanted to do this same concept with say a max of 40 steps, do I just need 5 relays?

Joel

 

[PRR]

> I use an excel sheet to make some test, but I'm too dumb and don't know how to use it to calculate all the 256 value

It must be possible to do binary operations in Excel. I don't know how. If it is like MS Basic, it could be a lot more difficult than it should be.

Am I correct in assuming that the In, Out, and Ground are as I've noted (in red)?

And that each relay is a 2-pole unit and switches two resistors at once?

By the way, I think you switched K8 and K7 in the lower-right.

What value do you like for "R"? 10K? That seems like a suitable trial value, because then 128R is 1.28Megs and R/128 is 78 ohms, practical values.

Then in the position shown, all relay contacts "up", the gain is 0 dB and the input impedance is 39 ohms (all the bottom resistors in parallel).

At the other extreme, all relay contacts "down", the gain is about 1/2 or about -6dB, the input impedance about 2.5 Megs.

Using binary logic, the "middle position" is with half the relays energized. If K1-K4 are energized, the top leg is about 250R||R which is essentially just R. The bottom leg is about R/60||R, which is about 1/60R. Input impedance is about R. Gain is about 1/60 or -35dB.

Or maybe I mis-read the binary logic, and the the "middle position" is the other half of the relays energized. If K5-K8 are energized, the top leg is about 16R||R which is essentially just R. The bottom leg is about R/16||R, which is about 1/16R. Input impedance is about R. Gain is about 1/16 or -24dB.

It seems odd that binary 00000000 gives 0dB, 11111111 gives -6dB, and 00001111 gives -35dB or 11110000 gives -24dB.

A 39 ohm input impedance is very awkward. If you assume R should be 1Meg, the input impedance only goes up to 3900 ohms, but you need 128Meg resistors. You can buy resistors marked over 100 Megs, but stray capacitance will leak signal around them like a much lower impedance.

I don't think it works the way you hoped.

 

[pilo]

mm You're right I do this schematic wrong...

of course K7 and K8 in the right bottom corner are switched...

and the the bottom switch are not in the right position too. I made this to fast I'm going to change it.

By the way it seems impossible to make a real log curve with it. I mean, each step decrease gain by 1db (or 2db, or any linear dB variations) and keeping the same value between in and gnd.

 

[mcs]

[quote author="pilo"]By the way it seems impossible to make a real log curve with it. I mean, each step decrease gain by 1db (or 2db, or any linear dB variations) and keeping the same value between in and gnd.[/quote]
If you need a circuit with in, out and gnd connections (like a standard volume pot), you can use this circuit: http://stiftsbogtrykkeriet.dk/~mcs/6bitatt.gif

It has a fixed input impedance and a fixed stepsize (1dB with the values shown I think). The only requirement is that the load impedance is fixed - or large compared to the attenuator impedance. The attenuator could be expanded to 8 relays if you really need 256 steps, instead of the 64 you get with 6 relays.

Best regards,

Mikkel C. Simonsen

 

[Moby]

Hi guys!
im jumped into DIY "digitally controlled stepped attenuator"...
And have some question about.... :roll:
I understood that "thing " is working in some logic way, but how empty AT89C4051 knows that logic? I dont see any software what will be used for driving relays in some logic way....or im missing something :?:
MCS, you mentioned that chip can be pre-programmed or you can store settings by buttons. Are you talking about ATMEL or EEPROM.
BTW, witch EEPROM we are talking about?
:thumb:

 

[mcs]

[quote author="Moby"]I understood that "thing " is working in some logic way, but how empty AT89C4051 knows that logic?[/quote]
It doesn't. You have to program the "logic" into the chip.

MCS, you mentioned that chip can be pre-programmed or you can store settings by buttons. Are you talking about ATMEL or EEPROM.

The basic software has to be programmed into the Atmel chip. If you know what taper you need, you can program that into the Atmel chip as well.

If you want a programmable taper instead, you can use an EEPROM to store the taper, using buttons to "input" the taper.

BTW, witch EEPROM we are talking about?

A cheap SPI EEPROM. The ones I use are called 95040 and are made by ST.

Best regards,

Mikkel C. Simonsen

 

[Moby]

Thanks Mikkel, now i know what to do.... :wink:

 

[pilo]

Hi Mikkel,
I want to port your relay stepped attenuator to the pic18f452 (for an intregration with MIOS).
So I would like to know what are the best value for log and/or anti-log curve?
I don't really know which value to "keep" : 0db, -1db...
Or can I use other value?

thanks

Pilo